The present invention relates to image recognition in detecting a moving object in an image, and particularly relates to a technology for realizing generation of a synthetic image according to a shooting state and detection of a moving object in an image in combination.
In conventional technologies in surrounding monitoring systems, there are techniques for detecting a moving object or an obstacle which involve hindrance for driving from images of vehicle surroundings shot by a camera installed to a vehicle and techniques for displaying a state of the vehicle surroundings as an image including the moving object or the obstacle.
Referring to an example of the techniques of moving object detection for monitoring vehicle surroundings, there is a technique in which a moving object approaching a vehicle is detected from images obtained by shooting vehicle surroundings with the use of optical flow (for example, Patent Documents 1 and 2). In Patent Document 1, for example, optical flow is calculated form images obtained by a camera installed so as to look rearward of a vehicle. Flow vectors having a magnitude equal to or larger than a predetermined threshold value and having the same direction as an approaching object are extracted. Then, the approaching object is discriminated on the basis of the flow vectors.
Referring to an example of the techniques of displaying a state of vehicle surroundings as an image, there is a technique of displaying a moving object detected from an input image shot by a camera, as a synthetic image. The synthetic image is generated and displayed as if it is shot from a position different from the camera position (for example, Patent Documents 3 to 5). In Patent Document 3, for example, with the use of a system shown in FIG. 16(a), a road region and a non-road region in which an obstacle is present are separated from an input image obtained by shooting vehicle surroundings as shown in FIG. 16(b). The road region is subjected to deformation processing to obtain an image viewed from above while the non-road region is subjected to expansion/contradiction processing to be in an appropriate size. Then, the obtained images are superimposed and displayed as a synthetic image as shown in FIG. 16(c).    Patent Document 1: Japanese Patent No. 3011566B    Patent Document 2: Japanese Patent Application Laid Open Publication No. 2004-56763A    Patent Document 3: Japanese Patent Application Laid Open Publication No. 07-186833A    Patent Document 4: Japanese Patent Application Laid Open Publication No. 06-333200A    Patent Document 5: Japanese Patent Application Laid Open Publication No. 11-78692A